KVM nVHE Hypervisor stack unwinder
v6 cover KVM nVHE Hypervisor stack unwinder
Hi all,
This is v6 of nVHE stacktrace support. The series is based on arm64 for-next/stacktrace.
The previous versions were posted at:
v5: https://lore.kernel.org/r/[email protected]/
v4: https://lore.kernel.org/r/[email protected]/
v3: https://lore.kernel.org/r/[email protected]/
v2: https://lore.kernel.org/r/[email protected]/
v1: https://lore.kernel.org/r/[email protected]/
This version mainly reorganizes the order of to patches into more coherent groups, along with addressing the other feedback in v5.
Patches 1 to 6 factor out the common unwinding logic into into arch/arm64/include/asm/stacktrace/common.h, for reuse by the nVHE unwinder. No functional change is introduced by these:
[01/17] arm64: stacktrace: Add shared header for common stack unwinding code [02/17] arm64: stacktrace: Factor out on_accessible_stack_common() [03/17] arm64: stacktrace: Factor out unwind_next_common() [04/17] arm64: stacktrace: Handle frame pointer from different address spaces [05/17] arm64: stacktrace: Factor out common unwind()
[06/17] arm64: stacktrace: Add description of stacktrace/common.h
Patches 7 to 11 implement stacktraces for conventional nVHE (non-protected mode):
[07/17] KVM: arm64: On stack overflow switch to hyp overflow_stack [08/17] KVM: arm64: Stub implementation of non-protected nVHE HYP stack
unwinder
[09/17] KVM: arm64: Prepare non-protected nVHE hypervisor stacktrace [10/17] KVM: arm64: Implement non-protected nVHE hyp stack unwinder [11/17] KVM: arm64: Introduce hyp_dump_backtrace()
Patches 12 to 17 implement stacktraces for protected-nVHE (pKVM):
[12/17] KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig [13/17] KVM: arm64: Allocate shared pKVM hyp stacktrace buffers [14/17] KVM: arm64: Stub implementation of pKVM HYP stack unwinder [15/17] KVM: arm64: Save protected-nVHE (pKVM) hyp stacktrace [16/17] KVM: arm64: Implement protected nVHE hyp stack unwinder [17/17] KVM: arm64: Introduce pkvm_dump_backtrace()
The updated cover letter is copied below for convenience.
Thanks,
Kalesh
============
KVM nVHE Stack unwinding.
===
nVHE has two modes of operation: protected (pKVM) and unprotected (conventional nVHE). Depending on the mode, a slightly different approach is used to dump the hypervisor stacktrace but the core unwinding logic remains the same.
Protected nVHE (pKVM) stacktraces
====
In protected nVHE mode, the host cannot directly access hypervisor memory.
The hypervisor stack unwinding happens in EL2 and is made accessible to the host via a shared buffer. Symbolizing and printing the stacktrace addresses is delegated to the host and happens in EL1.
Non-protected (Conventional) nVHE stacktraces
====
In non-protected mode, the host is able to directly access the hypervisor stack pages.
The hypervisor stack unwinding and dumping of the stacktrace is performed by the host in EL1, as this avoids the memory overhead of setting up shared buffers between the host and hypervisor.
Resuing the Core Unwinding Logic
====
Since the hypervisor cannot link against the kernel code in proteced mode. The common stack unwinding code is moved to a shared header to allow reuse in the nVHE hypervisor.
Reducing the memory footprint
====
In this version the below steps were taken to reduce the memory usage of nVHE stack unwinding:
1) The nVHE overflow stack is reduced from PAGE_SIZE to 4KB; benificial for configurations with non 4KB pages (16KB or 64KB pages).
2) In protected nVHE mode (pKVM), the shared stacktrace buffers with the host are reduced from PAGE_SIZE to the minimum size required.
3) In systems other than Android, conventional nVHE makes up the vast majority of use case. So the pKVM stack tracing is disabled by default (!CONFIG_PROTECTED_NVHE_STACKTRACE), which avoid the memory usage for setting up shared buffers.
4) In non-protected nVHE mode (conventional nVHE), the stack unwinding is done directly in EL1 by the host and no shared buffers with the hypervisor are needed.
Sample Output
====
The below shows an example output from a simple stack overflow test:
[ 111.623091] kvm [367]: nVHE call trace: [ 111.623215] kvm [367]: [<ffff8000090a6570>] __kvm_nvhe_hyp_panic+0xac/0xf8 [ 111.623448] kvm [367]: [<ffff8000090a65cc>] __kvm_nvhe_hyp_panic_bad_stack+0x10/0x10 [ 111.623642] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 . . . [ 111.640366] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 111.640467] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 111.640574] kvm [367]: [<ffff8000090a5de4>] __kvm_nvhe___kvm_vcpu_run+0x30/0x40c [ 111.640676] kvm [367]: [<ffff8000090a8b64>] __kvm_nvhe_handle___kvm_vcpu_run+0x30/0x48 [ 111.640778] kvm [367]: [<ffff8000090a88b8>] __kvm_nvhe_handle_trap+0xc4/0x128 [ 111.640880] kvm [367]: [<ffff8000090a7864>] __kvm_nvhe___host_exit+0x64/0x64 [ 111.640996] kvm [367]: ---[ end nVHE call trace ]---
============
Kalesh Singh (17):
arm64: stacktrace: Add shared header for common stack unwinding code
arm64: stacktrace: Factor out on_accessible_stack_common()
arm64: stacktrace: Factor out unwind_next_common()
arm64: stacktrace: Handle frame pointer from different address spaces
arm64: stacktrace: Factor out common unwind()
arm64: stacktrace: Add description of stacktrace/common.h
KVM: arm64: On stack overflow switch to hyp overflow_stack
KVM: arm64: Stub implementation of non-protected nVHE HYP stack
unwinder
KVM: arm64: Prepare non-protected nVHE hypervisor stacktrace
KVM: arm64: Implement non-protected nVHE hyp stack unwinder
KVM: arm64: Introduce hyp_dump_backtrace()
KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig
KVM: arm64: Allocate shared pKVM hyp stacktrace buffers
KVM: arm64: Stub implementation of pKVM HYP stack unwinder
KVM: arm64: Save protected-nVHE (pKVM) hyp stacktrace
KVM: arm64: Implement protected nVHE hyp stack unwinder
KVM: arm64: Introduce pkvm_dump_backtrace()
arch/arm64/include/asm/kvm_asm.h | 16 ++ arch/arm64/include/asm/memory.h | 8 + arch/arm64/include/asm/stacktrace.h | 92 +++++---- arch/arm64/include/asm/stacktrace/common.h | 230 +++++++++++++++++++++ arch/arm64/include/asm/stacktrace/nvhe.h | 176 ++++++++++++++++ arch/arm64/kernel/stacktrace.c | 157 -------------- arch/arm64/kvm/Kconfig | 15 ++ arch/arm64/kvm/arm.c | 2 +- arch/arm64/kvm/handle_exit.c | 102 +++++++++ arch/arm64/kvm/hyp/nvhe/Makefile | 2 +- arch/arm64/kvm/hyp/nvhe/host.S | 9 +- arch/arm64/kvm/hyp/nvhe/stacktrace.c | 109 ++++++++++ arch/arm64/kvm/hyp/nvhe/switch.c | 6 + 13 files changed, 720 insertions(+), 204 deletions(-) create mode 100644 arch/arm64/include/asm/stacktrace/common.h create mode 100644 arch/arm64/include/asm/stacktrace/nvhe.h create mode 100644 arch/arm64/kvm/hyp/nvhe/stacktrace.c
base-commit: 82a592c13b0aeff94d84d54183dae0b26384c95f
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 1/17 arm64: stacktrace: Add shared header for common stack unwinding code
In order to reuse the arm64 stack unwinding logic for the nVHE hypervisor stack, move the common code to a shared header (arch/arm64/include/asm/stacktrace/common.h).
The nVHE hypervisor cannot safely link against kernel code, so we make use of the shared header to avoid duplicated logic later in this series.
---
Changes in v6:
- Add Fuad's Tested-by tag
Changes in v5:
- Add Reviewed-by tags from Mark Brown and Fuad
arch/arm64/include/asm/stacktrace.h | 35 +------ arch/arm64/include/asm/stacktrace/common.h | 105 +++++++++++++++++++++ arch/arm64/kernel/stacktrace.c | 57 ----------- 3 files changed, 106 insertions(+), 91 deletions(-) create mode 100644 arch/arm64/include/asm/stacktrace/common.h
arch/arm64/include/asm/stacktrace.h +1 -34
@@ -8,52 +8,19 @@#include <linux/percpu.h> #include <linux/sched.h> #include <linux/sched/task_stack.h> -#include <linux/types.h> #include <linux/llist.h> #include <asm/memory.h> #include <asm/ptrace.h> #include <asm/sdei.h> -enum stack_type { -STACK_TYPE_UNKNOWN , -STACK_TYPE_TASK , -STACK_TYPE_IRQ , -STACK_TYPE_OVERFLOW , -STACK_TYPE_SDEI_NORMAL , -STACK_TYPE_SDEI_CRITICAL , - __NR_STACK_TYPES -}; - -struct stack_info { -unsigned long low; -unsigned long high; -enum stack_type type; -}; +#include <asm/stacktrace/common.h> extern void dump_backtrace (struct pt_regs *regs,struct task_struct *tsk,const char *loglvl); DECLARE_PER_CPU(unsigned long *, irq_stack_ptr); -static inline bool on_stack (unsigned long sp,unsigned long size, -unsigned long low,unsigned long high, -enum stack_type type,struct stack_info *info) -{ -if (!low) -return false; - -if (sp < low || sp + size < sp || sp + size > high) -return false; - -if (info) { - info->low = low; - info->high = high; - info->type = type; - } -return true; -} -static inline bool on_irq_stack (unsigned long sp,unsigned long size,struct stack_info *info) {
arch/arm64/include/asm/stacktrace/common.h new file +105 -0
105 lines not on this page · raw message
arch/arm64/kernel/stacktrace.c +0 -57
63 lines not on this page · raw message
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 2/17 arm64: stacktrace: Factor out on_accessible_stack_common()
In reply to Kalesh SinghMove common on_accessible_stack checks to stacktrace/common.h. This is used in the implementation of the nVHE hypervisor unwinder later in this series.
---
Changes in v6:
- Add Fuad's Tested-by tag
Changes in v5:
- Add Reviewed-by tags from Mark Brown and Fuad
- Remove random whitespace change, per Mark Brown
arch/arm64/include/asm/stacktrace.h | 6 ++---- arch/arm64/include/asm/stacktrace/common.h | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 4 deletions(-)
arch/arm64/include/asm/stacktrace.h +2 -4
@@ -65,8 +65,8 @@ static inline bool on_accessible_stack(const struct task_struct *tsk, unsigned long sp, unsigned long size, struct stack_info *info) { - if (info) - info->type = STACK_TYPE_UNKNOWN; + if (on_accessible_stack_common(tsk, sp, size, info)) + return true;if (on_task_stack (tsk, sp, size, info))return true;
@@ -74,8 +74,6 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,return false;if (on_irq_stack (sp, size, info))return true; -if (on_overflow_stack (sp, size, info)) -return true;if (on_sdei_stack (sp, size, info))return true;
arch/arm64/include/asm/stacktrace/common.h +18 -0
@@ -62,6 +62,9 @@ struct unwind_state {struct task_struct *task; }; +static inline bool on_overflow_stack (unsigned long sp,unsigned long size, +struct stack_info *info); +static inline bool on_stack (unsigned long sp,unsigned long size,unsigned long low,unsigned long high,enum stack_type type,struct stack_info *info)
@@ -80,6 +83,21 @@ static inline bool on_stack(unsigned long sp, unsigned long size,return true; } +static inline bool on_accessible_stack_common (const struct task_struct *tsk, +unsigned long sp, +unsigned long size, +struct stack_info *info) +{ +if (info) + info->type =STACK_TYPE_UNKNOWN ; + +/* +* Both the kernel and nvhe hypervisor make use of +* an overflow_stack +*/ +return on_overflow_stack (sp, size, info); +} +static inline void unwind_init_common (struct unwind_state *state,struct task_struct *task) {
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v6 02/17] arm64: stacktrace: Factor out on_accessible_stack_common()
Quoted 68 lines · Kalesh Singh
Move common
on_accessible_stackchecks tostacktrace/common.h. This is used in the implementation of the nVHE hypervisor unwinder later in this series.Signed-off-by Kalesh Singh <[email protected]>Reviewed-by Fuad Tabba <[email protected]>Reviewed-by Mark Brown <[email protected]>Tested-by Fuad Tabba <[email protected]>---
Changes in v6:
- Add Fuad's Tested-by tagChanges in v5:
- Add Reviewed-by tags from Mark Brown and Fuad
- Remove random whitespace change, per Mark Brownarch/arm64/include/asm/stacktrace.h | 6 ++---- arch/arm64/include/asm/stacktrace/common.h | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 4 deletions(-)arch/arm64/include/asm/stacktrace.h +2
-4@@ -65,8 +65,8 @@ static inline bool on_accessible_stack(const struct task_struct *tsk, unsigned long sp, unsigned long size, struct stack_info *info) { - if (info) - info->type = STACK_TYPE_UNKNOWN; + if (on_accessible_stack_common(tsk, sp, size, info)) + return true;if (on_task_stack (tsk, sp, size, info))return true;@@ -74,8 +74,6 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,return false;if (on_irq_stack (sp, size, info))return true; -if (on_overflow_stack (sp, size, info)) -return true;if (on_sdei_stack (sp, size, info))return true;arch/arm64/include/asm/stacktrace/common.h +7
-0@@ -62,6 +62,9 @@ struct unwind_state {struct task_struct *task; }; +static inline bool on_overflow_stack (unsigned long sp,unsigned long size, +struct stack_info *info); +static inline bool on_stack (unsigned long sp,unsigned long size,unsigned long low,unsigned long high,enum stack_type type,struct stack_info *info)@@ -80,6 +83,21 @@ static inline bool on_stack(unsigned long sp, unsigned long size,return true; } +static inline bool on_accessible_stack_common (const struct task_struct *tsk, +unsigned long sp, +unsigned long size, +struct stack_info *info)
arch/arm64/include/asm/stacktrace/common.h static inline bool on_stack(unsigned long sp, unsigned long size, · line 99
Quoted 10 lines
arch/arm64/include/asm/stacktrace/common.h +10
-0+{ +if (info) + info->type =STACK_TYPE_UNKNOWN ; + +/* +* Both the kernel and nvhe hypervisor make use of +* an overflow_stack +*/ +return on_overflow_stack (sp, size, info); +}
on_accessible_stack has the following comment:
/*
* We can only safely access per-cpu stacks from current in a non-preemptible
* context.
*/
With this change, I don't think we satisfy this requirement anymore, as we're checking the overflow stack *before* the preemptible check, which is a big change in behaviour.
The hypervisor doesn't have this requirement: the unwinding is either done out of context (nVHE, where EL1 unwinds EL2) or in a non-preemptible section (pKVM, where the whole thing is non-preemptible). But the kernel is usually preemptible, so this patch needs fixing.
I'll see if I can address it locally (I'm currently moving things around, stay tuned).
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 02/17] arm64: stacktrace: Factor out on_accessible_stack_common()
Quoted 93 lines · Marc Zyngier
Quoted 68 lines · Kalesh Singh
Move common
on_accessible_stackchecks tostacktrace/common.h. This is used in the implementation of the nVHE hypervisor unwinder later in this series.Signed-off-by Kalesh Singh <[email protected]>Reviewed-by Fuad Tabba <[email protected]>Reviewed-by Mark Brown <[email protected]>Tested-by Fuad Tabba <[email protected]>---
Changes in v6:
- Add Fuad's Tested-by tagChanges in v5:
- Add Reviewed-by tags from Mark Brown and Fuad
- Remove random whitespace change, per Mark Brownarch/arm64/include/asm/stacktrace.h | 6 ++---- arch/arm64/include/asm/stacktrace/common.h | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 4 deletions(-)arch/arm64/include/asm/stacktrace.h +2
-4@@ -65,8 +65,8 @@ static inline bool on_accessible_stack(const struct task_struct *tsk, unsigned long sp, unsigned long size, struct stack_info *info) { - if (info) - info->type = STACK_TYPE_UNKNOWN; + if (on_accessible_stack_common(tsk, sp, size, info)) + return true;if (on_task_stack (tsk, sp, size, info))return true;@@ -74,8 +74,6 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,return false;if (on_irq_stack (sp, size, info))return true; -if (on_overflow_stack (sp, size, info)) -return true;if (on_sdei_stack (sp, size, info))return true;arch/arm64/include/asm/stacktrace/common.h +7
-0@@ -62,6 +62,9 @@ struct unwind_state {struct task_struct *task; }; +static inline bool on_overflow_stack (unsigned long sp,unsigned long size, +struct stack_info *info); +static inline bool on_stack (unsigned long sp,unsigned long size,unsigned long low,unsigned long high,enum stack_type type,struct stack_info *info)@@ -80,6 +83,21 @@ static inline bool on_stack(unsigned long sp, unsigned long size,return true; } +static inline bool on_accessible_stack_common (const struct task_struct *tsk, +unsigned long sp, +unsigned long size, +struct stack_info *info)arch/arm64/include/asm/stacktrace/common.h static inline bool on_stack(unsigned long sp, unsigned long size, · line 99
Quoted 10 lines
arch/arm64/include/asm/stacktrace/common.h +10
-0+{ +if (info) + info->type =STACK_TYPE_UNKNOWN ; + +/* +* Both the kernel and nvhe hypervisor make use of +* an overflow_stack +*/ +return on_overflow_stack (sp, size, info); +}
on_accessible_stackhas the following comment:/*
* We can only safely access per-cpu stacks from current in a non-preemptible
* context.
*/With this change, I don't think we satisfy this requirement anymore, as we're checking the overflow stack *before* the preemptible check, which is a big change in behaviour.
Quoted 8 lines
The hypervisor doesn't have this requirement: the unwinding is either done out of context (nVHE, where EL1 unwinds EL2) or in a non-preemptible section (pKVM, where the whole thing is non-preemptible). But the kernel is usually preemptible, so this patch needs fixing.
I'll see if I can address it locally (I'm currently moving things around, stay tuned).
Hi Marc,
You are right it changes the order of the checks. I think the simplest way then is to drop this patch, and replace uses of on_accessible _stack_common().
Thanks,
Kalesh
Quoted 7 lines
Thanks,
M.--
Without deviation from the norm, progress is not possible.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 3/17 arm64: stacktrace: Factor out unwind_next_common()
In reply to Kalesh SinghMove common unwind_next logic to stacktrace/common.h. This allows reusing the code in the implementation the nVHE hypervisor stack unwinder, later in this series.
---
Changes in v6:
- Add Fuad's Tested-by tag
Changes in v5:
- Add Reviewed-by tags from Mark Brown and Fuad
arch/arm64/include/asm/stacktrace/common.h | 50 ++++++++++++++++++++++ arch/arm64/kernel/stacktrace.c | 41 ++---------------- 2 files changed, 54 insertions(+), 37 deletions(-)
arch/arm64/include/asm/stacktrace/common.h +50 -0
@@ -65,6 +65,10 @@ struct unwind_state {static inline bool on_overflow_stack (unsigned long sp,unsigned long size,struct stack_info *info); +static inline bool on_accessible_stack (const struct task_struct *tsk, +unsigned long sp,unsigned long size, +struct stack_info *info); +static inline bool on_stack (unsigned long sp,unsigned long size,unsigned long low,unsigned long high,enum stack_type type,struct stack_info *info)
@@ -120,4 +124,50 @@ static inline void unwind_init_common(struct unwind_state *state, state->prev_type =STACK_TYPE_UNKNOWN ; } +static inline int unwind_next_common (struct unwind_state *state, +struct stack_info *info) +{ +struct task_struct *tsk = state->task; +unsigned long fp = state->fp; + +if (fp &0x7 ) +return -EINVAL ; + +if (!on_accessible_stack (tsk, fp,16 , info)) +return -EINVAL ; + +if (test_bit (info->type, state->stacks_done)) +return -EINVAL ; + +/* +* As stacks grow downward, any valid record on the same stack must be +* at a strictly higher address than the prior record. +* +* Stacks can nest in several valid orders, e.g. +* +* TASK -> IRQ -> OVERFLOW -> SDEI_NORMAL +* TASK -> SDEI_NORMAL -> SDEI_CRITICAL -> OVERFLOW +* +* ... but the nesting itself is strict. Once we transition from one +* stack to another, it's never valid to unwind back to that first +* stack. +*/ +if (info->type == state->prev_type) { +if (fp <= state->prev_fp) +return -EINVAL ; + }else { +__set_bit (state->prev_type, state->stacks_done); + } + +/* +* Record this frame record's values and location. The prev_fp and +* prev_type are only meaningful to the next unwind_next() invocation. +*/ + state->fp =READ_ONCE (*(unsigned long *)(fp)); + state->pc =READ_ONCE (*(unsigned long *)(fp +8 )); + state->prev_fp = fp; + state->prev_type = info->type; + +return 0 ; +}#endif /* __ASM_STACKTRACE_COMMON_H */
arch/arm64/kernel/stacktrace.c +4 -37
52 lines not on this page · raw message
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 4/17 arm64: stacktrace: Handle frame pointer from different address spaces
In reply to Kalesh SinghThe unwinder code is made reusable so that it can be used to unwind various types of stacks. One usecase is unwinding the nVHE hyp stack from the host (EL1) in non-protected mode. This means that the unwinder must be able to translate HYP stack addresses to kernel addresses.
Add a callback (stack_trace_translate_fp_fn) to allow specifying the translation function.
---
Changes in v6:
- Fix typo in comment, per Fuad
- Add Fuad’s Reviewed-by tag and Tested-by tags
Changes in v5:
- Fix typo in commit text, per Fuad
- Update unwind_next_common() to not have side effects on failure, per Fuad
- Use regular comment instead of doc comments, per Fuad
arch/arm64/include/asm/stacktrace/common.h | 29 +++++++++++++++++++--- arch/arm64/kernel/stacktrace.c | 2 +- 2 files changed, 26 insertions(+), 5 deletions(-)
arch/arm64/include/asm/stacktrace/common.h +25 -4
@@ -124,11 +124,25 @@ static inline void unwind_init_common(struct unwind_state *state, state->prev_type =STACK_TYPE_UNKNOWN ; } +/* +* stack_trace_translate_fp_fn() - Translates a non-kernel frame pointer to +* a kernel address. +* +* @fp: the frame pointer to be updated to its kernel address. +* @type: the stack type associated with frame pointer @fp +* +* Returns true and success and @fp is updated to the corresponding +* kernel virtual address; otherwise returns false. +*/ +typedef bool (*stack_trace_translate_fp_fn )(unsigned long *fp, +enum stack_type type); +static inline int unwind_next_common (struct unwind_state *state, -struct stack_info *info) +struct stack_info *info, +stack_trace_translate_fp_fn translate_fp) { +unsigned long fp = state->fp, kern_fp = fp;struct task_struct *tsk = state->task; -unsigned long fp = state->fp;if (fp &0x7 )return -EINVAL ;
@@ -139,6 +153,13 @@ static inline int unwind_next_common(struct unwind_state *state,if (test_bit (info->type, state->stacks_done))return -EINVAL ; +/* +* If fp is not from the current address space perform the necessary +* translation before dereferencing it to get the next fp. +*/ +if (translate_fp && !translate_fp (&kern_fp, info->type)) +return -EINVAL ; +/* * As stacks grow downward, any valid record on the same stack must be * at a strictly higher address than the prior record.
@@ -163,8 +184,8 @@ static inline int unwind_next_common(struct unwind_state *state,* Record this frame record's values and location. The prev_fp and * prev_type are only meaningful to the next unwind_next() invocation. */ - state->fp =READ_ONCE (*(unsigned long *)(fp)); - state->pc =READ_ONCE (*(unsigned long *)(fp +8 )); + state->fp =READ_ONCE (*(unsigned long *)(kern_fp)); + state->pc =READ_ONCE (*(unsigned long *)(kern_fp +8 )); state->prev_fp = fp; state->prev_type = info->type;
arch/arm64/kernel/stacktrace.c +1 -1
@@ -87,7 +87,7 @@ static int notrace unwind_next(struct unwind_state *state)if (fp == (unsigned long )task_pt_regs (tsk)->stackframe)return -ENOENT ; - err =unwind_next_common (state, &info); + err =unwind_next_common (state, &info,NULL );if (err)return err;
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 04/17] arm64: stacktrace: Handle frame pointer from different address spaces
Quoted 8 lines · Kalesh Singh
The unwinder code is made reusable so that it can be used to unwind various types of stacks. One usecase is unwinding the nVHE hyp stack from the host (EL1) in non-protected mode. This means that the unwinder must be able to translate HYP stack addresses to kernel addresses.
Add a callback (
stack_trace_translate_fp_fn) to allow specifying the translation function.
with or without one very minor thing:
arch/arm64/include/asm/stacktrace/common.h static inline void unwind_init_common(struct unwind_state *state, · line 144
Quoted 6 lines
static inline int unwind_next_common(struct
unwind_state*state,arch/arm64/include/asm/stacktrace/common.h +3
-1- struct stack_info *info) + struct stack_info *info, + stack_trace_translate_fp_fn translate_fp) { +unsigned long fp = state->fp, kern_fp = fp;
As a coding style nit I don't love having multiple assignments on a single line especially as part of declarations.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 04/17] arm64: stacktrace: Handle frame pointer from different address spaces
Quoted 22 lines · Mark Brown
Quoted 8 lines · Kalesh Singh
The unwinder code is made reusable so that it can be used to unwind various types of stacks. One usecase is unwinding the nVHE hyp stack from the host (EL1) in non-protected mode. This means that the unwinder must be able to translate HYP stack addresses to kernel addresses.
Add a callback (
stack_trace_translate_fp_fn) to allow specifying the translation function.Reviewed-by Mark Brown <[email protected]>with or without one very minor thing:
arch/arm64/include/asm/stacktrace/common.h static inline void unwind_init_common(struct unwind_state *state, · line 144
Quoted 6 lines
static inline int unwind_next_common(struct
unwind_state*state,arch/arm64/include/asm/stacktrace/common.h +3
-1- struct stack_info *info) + struct stack_info *info, + stack_trace_translate_fp_fn translate_fp) { +unsigned long fp = state->fp, kern_fp = fp;
Quoted 2 lines
As a coding style nit I don't love having multiple assignments on a single line especially as part of declarations.
Hi Mark,
Thanks for the reviews. I'll update this if a respin is needed.
--Kalesh
PATCH v6 5/17 arm64: stacktrace: Factor out common unwind()
In reply to Kalesh SinghMove unwind() to stacktrace/common.h, and as a result the kernel unwind_next() to asm/stacktrace.h. This allow reusing unwind() in the implementation of the nVHE HYP stack unwinder, later in the series.
---
Changes in v6:
- Add Mark Brown’s Reviewed-by tag
- Add Fuad's Tested-by tag
Changes in v5:
- Add Reviewed-by tag from Fuad
arch/arm64/include/asm/stacktrace.h | 51 ++++++++++++++++ arch/arm64/include/asm/stacktrace/common.h | 19 ++++++ arch/arm64/kernel/stacktrace.c | 67 ---------------------- 3 files changed, 70 insertions(+), 67 deletions(-)
arch/arm64/include/asm/stacktrace.h +51 -0
@@ -11,6 +11,7 @@#include <linux/llist.h> #include <asm/memory.h> +#include <asm/pointer_auth.h> #include <asm/ptrace.h> #include <asm/sdei.h>
@@ -80,4 +81,54 @@ static inline bool on_accessible_stack(const struct task_struct *tsk, return false; } +/* +* Unwind from one frame record (A) to the next frame record (B). +* +* We terminate early if the location of B indicates a malformed chain of frame +* records (e.g. a cycle), determined based on the location and fp value of A +* and the location (but not the fp value) of B. +*/ +static inline int notrace unwind_next(struct unwind_state *state) +{ +struct task_struct *tsk = state->task; +unsigned long fp = state->fp; +struct stack_info info; +int err; + +/* Final frame; nothing to unwind */ +if (fp == (unsigned long )task_pt_regs (tsk)->stackframe) +return -ENOENT ; + + err =unwind_next_common (state, &info,NULL ); +if (err) +return err; + + state->pc =ptrauth_strip_insn_pac (state->pc); + +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +if (tsk->ret_stack && + (state->pc == (unsigned long )return_to_handler)) { +unsigned long orig_pc; +/* +* This is a case where function graph tracer has +* modified a return address (LR) in a stack frame +* to hook a function return. +* So replace it to an original value. +*/ + orig_pc =ftrace_graph_ret_addr (tsk,NULL , state->pc, + (void *)state->fp); +if (WARN_ON_ONCE (state->pc == orig_pc)) +return -EINVAL ; + state->pc = orig_pc; + } +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ +#ifdef CONFIG_KRETPROBES +if (is_kretprobe_trampoline (state->pc)) + state->pc =kretprobe_find_ret_addr (tsk, (void *)state->fp, &state->kr_cur); +#endif + +return 0 ; +} +NOKPROBE_SYMBOL (unwind_next); +#endif /* __ASM_STACKTRACE_H */
arch/arm64/include/asm/stacktrace/common.h +19 -0
@@ -9,6 +9,7 @@#include <linux/bitmap.h> #include <linux/bitops.h> +#include <linux/kprobes.h> #include <linux/types.h> enum stack_type {
@@ -69,6 +70,8 @@ static inline bool on_accessible_stack(const struct task_struct *tsk, unsigned long sp, unsigned long size, struct stack_info *info); +static inline int unwind_next (struct unwind_state *state); +static inline bool on_stack (unsigned long sp,unsigned long size,unsigned long low,unsigned long high,enum stack_type type,struct stack_info *info)
@@ -191,4 +194,20 @@ static inline int unwind_next_common(struct unwind_state *state,return 0 ; } + +static inline void notrace unwind(struct unwind_state *state, +stack_trace_consume_fn consume_entry, +void *cookie) +{ +while (1 ) { +int ret; + +if (!consume_entry (cookie, state->pc)) +break ; + ret =unwind_next (state); +if (ret <0 ) +break ; + } +} +NOKPROBE_SYMBOL (unwind);#endif /* __ASM_STACKTRACE_COMMON_H */
arch/arm64/kernel/stacktrace.c +0 -67
85 lines not on this page · raw message
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 6/17 arm64: stacktrace: Add description of stacktrace/common.h
In reply to Kalesh SinghAdd brief description on how to use stacktrace/common.h to implement a stack unwinder.
---
Changes in v6:
- Add Fuad’s Reviewed-by tag
Changes in v5:
- Add short description of each required function, per Fuad and Marc
arch/arm64/include/asm/stacktrace/common.h | 15 +++++++++++++++ 1 file changed, 15 insertions(+)
arch/arm64/include/asm/stacktrace/common.h +15 -0
@@ -2,6 +2,21 @@/* * Common arm64 stack unwinder code. * +* To implement a new arm64 stack unwinder: +* 1) Include this header +* +* 2) Provide implementations for the following functions: +* on_overflow_stack(): Returns true if SP is on the overflow +* stack. +* on_accessible_stack(): Returns true is SP is on any accessible +* stack. +* unwind_next(): Performs validation checks on the frame +* pointer, and transitions unwind_state +* to the next frame. +* +* See: arch/arm64/include/asm/stacktrace.h for reference +* implementations. +* * Copyright (C) 2012 ARM Ltd. */ #ifndef __ASM_STACKTRACE_COMMON_H
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 06/17] arm64: stacktrace: Add description of stacktrace/common.h
Quoted 2 lines · Kalesh Singh
Add brief description on how to use
stacktrace/common.hto implement a stack unwinder.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 7/17 KVM: arm64: On stack overflow switch to hyp overflow_stack
In reply to Kalesh SinghOn hyp stack overflow switch to 16-byte aligned secondary stack. This provides us stack space to better handle overflows; and is used in a subsequent patch to dump the hypervisor stacktrace.
---
Changes in v6:
- Add Fuad's Tested-by tag
Changes in v5:
- Add Reviewed-by tag from Fuad
arch/arm64/kvm/hyp/nvhe/Makefile | 2 +- arch/arm64/kvm/hyp/nvhe/host.S | 9 ++------- arch/arm64/kvm/hyp/nvhe/stacktrace.c | 11 +++++++++++ 3 files changed, 14 insertions(+), 8 deletions(-) create mode 100644 arch/arm64/kvm/hyp/nvhe/stacktrace.c
arch/arm64/kvm/hyp/nvhe/Makefile +1 -1
@@ -14,7 +14,7 @@ lib-objs := $(addprefix ../../../lib/, $(lib-objs)) obj-y := timer-sr.o sysreg-sr.o debug-sr.o switch.o tlb.o hyp-init.o host.o \ hyp-main.o hyp-smp.o psci-relay.o early_alloc.o page_alloc.o \ - cache.o setup.o mm.o mem_protect.o sys_regs.o pkvm.o + cache.o setup.o mm.o mem_protect.o sys_regs.o pkvm.o stacktrace.o obj-y += ../vgic-v3-sr.o ../aarch32.o ../vgic-v2-cpuif-proxy.o ../entry.o \ ../fpsimd.o ../hyp-entry.o ../exception.o ../pgtable.o obj-$(CONFIG_DEBUG_LIST) += list_debug.o
arch/arm64/kvm/hyp/nvhe/host.S +2 -7
@@ -177,13 +177,8 @@ SYM_FUNC_END(__host_hvc) b hyp_panic .L__hyp_sp_overflow\@: - /* - * Reset SP to the top of the stack, to allow handling the hyp_panic. - * This corrupts the stack but is ok, since we won't be attempting - * any unwinding here. - */ - ldr_this_cpu x0, kvm_init_params + NVHE_INIT_STACK_HYP_VA, x1 - mov sp, x0 + /* Switch to the overflow stack */ + adr_this_cpu sp, overflow_stack + OVERFLOW_STACK_SIZE, x0 b hyp_panic_bad_stack ASM_BUG()
arch/arm64/kvm/hyp/nvhe/stacktrace.c new file +11 -0
@@ -0,0 +1,11 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* KVM nVHE hypervisor stack tracing support. +* +* Copyright (C) 2022 Google LLC +*/ +#include <asm/memory.h> +#include <asm/percpu.h> + +DEFINE_PER_CPU(unsigned long [OVERFLOW_STACK_SIZE/sizeof(long)], overflow_stack) +__aligned(16) ;
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 8/17 KVM: arm64: Stub implementation of non-protected nVHE HYP stack unwinder
In reply to Kalesh SinghAdd stub implementations of non-protected nVHE stack unwinder, for building. These are implemented later in this series.
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Mark unwind_next() as inline, per Marc
- Comment !__KVM_NVHE_HYPERVISOR__ unwinder path, per Marc
arch/arm64/include/asm/stacktrace/nvhe.h | 47 ++++++++++++++++++++++++ 1 file changed, 47 insertions(+) create mode 100644 arch/arm64/include/asm/stacktrace/nvhe.h
arch/arm64/include/asm/stacktrace/nvhe.h new file +47 -0
@@ -0,0 +1,47 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* KVM nVHE hypervisor stack tracing support. +* +* The unwinder implementation depends on the nVHE mode: +* +* 1) Non-protected nVHE mode - the host can directly access the +* HYP stack pages and unwind the HYP stack in EL1. This saves having +* to allocate shared buffers for the host to read the unwinded +* stacktrace. +* +* Copyright (C) 2022 Google LLC +*/ +#ifndef __ASM_STACKTRACE_NVHE_H +#define __ASM_STACKTRACE_NVHE_H + +#include <asm/stacktrace/common.h> + +static inline bool on_accessible_stack (const struct task_struct *tsk, +unsigned long sp,unsigned long size, +struct stack_info *info) +{ +return false; +} + +#ifndef __KVM_NVHE_HYPERVISOR__ +/* +* Conventional (non-protected) nVHE HYP stack unwinder +* +* In non-protected mode, the unwinding is done from kernel proper context +* (by the host in EL1). +*/ + +static inline bool on_overflow_stack (unsigned long sp,unsigned long size, +struct stack_info *info) +{ +return false; +} + +static inline int notrace unwind_next(struct unwind_state *state) +{ +return 0 ; +} +NOKPROBE_SYMBOL (unwind_next); + +#endif /* !__KVM_NVHE_HYPERVISOR__ */ +#endif /* __ASM_STACKTRACE_NVHE_H */
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 9/17 KVM: arm64: Prepare non-protected nVHE hypervisor stacktrace
In reply to Kalesh SinghIn non-protected nVHE mode (non-pKVM) the host can directly access hypervisor memory; and unwinding of the hypervisor stacktrace is done from EL1 to save on memory for shared buffers.
To unwind the hypervisor stack from EL1 the host needs to know the starting point for the unwind and information that will allow it to translate hypervisor stack addresses to the corresponding kernel addresses. This patch sets up this book keeping. It is made use of later in the series.
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Use regular comments instead of doc comments, per Fuad
arch/arm64/include/asm/kvm_asm.h | 16 +++++++++++ arch/arm64/kvm/hyp/nvhe/stacktrace.c | 41 ++++++++++++++++++++++++++++ arch/arm64/kvm/hyp/nvhe/switch.c | 6 ++++ 3 files changed, 63 insertions(+)
arch/arm64/include/asm/kvm_asm.h +16 -0
@@ -176,6 +176,22 @@ struct kvm_nvhe_init_params {unsigned long vtcr; }; +/* +* Used by the host in EL1 to dump the nVHE hypervisor backtrace on +* hyp_panic() in non-protected mode. +* +* @stack_base: hyp VA of the hyp_stack base. +* @overflow_stack_base: hyp VA of the hyp_overflow_stack base. +* @fp: hyp FP where the backtrace begins. +* @pc: hyp PC where the backtrace begins. +*/ +struct kvm_nvhe_stacktrace_info { +unsigned long stack_base; +unsigned long overflow_stack_base; +unsigned long fp; +unsigned long pc; +}; +/* Translate a kernel address @ptr into its equivalent linear mapping */ #define kvm_ksym_ref (ptr) \ ({ \
arch/arm64/kvm/hyp/nvhe/stacktrace.c +41 -0
49 lines not on this page · raw message
arch/arm64/kvm/hyp/nvhe/switch.c +6 -0
@@ -34,6 +34,8 @@ DEFINE_PER_CPU(struct kvm_host_data, kvm_host_data); DEFINE_PER_CPU(struct kvm_cpu_context, kvm_hyp_ctxt); DEFINE_PER_CPU(unsigned long, kvm_hyp_vector); +extern void kvm_nvhe_prepare_backtrace (unsigned long fp,unsigned long pc); +static void __activate_traps (struct kvm_vcpu *vcpu) {u64 val;
@@ -375,6 +377,10 @@ asmlinkage void __noreturn hyp_panic(void)__sysreg_restore_state_nvhe (host_ctxt); } +/* Prepare to dump kvm nvhe hyp stacktrace */ +kvm_nvhe_prepare_backtrace ((unsigned long )__builtin_frame_address (0 ), + _THIS_IP_); +__hyp_do_panic (host_ctxt, spsr, elr, par);unreachable (); }
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
kernel test robot
build botRe: [PATCH v6 09/17] KVM: arm64: Prepare non-protected nVHE hypervisor stacktrace
Thank you for the patch! Perhaps something to improve:
Hi Kalesh,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on 82a592c13b0aeff94d84d54183dae0b26384c95f]
url: https://github.com/intel-lab-lkp/linux/commits/Kalesh-Singh/KVM-nVHE-Hypervisor-stack-unwinder/20220726-154355 base: 82a592c13b0aeff94d84d54183dae0b26384c95f config: arm64-buildonly-randconfig-r006-20220724 (https://download.01.org/0day-ci/archive/20220727/[email protected]/config) compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project 83882606dbd7ffb0bdd3460356202d97705809c8)
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# install arm64 cross compiling tool for clang build
# apt-get install binutils-aarch64-linux-gnu
# https://github.com/intel-lab-lkp/linux/commit/adb3292a8f22d642d40a5f2f649592a1961fa7da
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Kalesh-Singh/KVM-nVHE-Hypervisor-stack-unwinder/20220726-154355 git checkout adb3292a8f22d642d40a5f2f649592a1961fa7da
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=arm64 SHELL=/bin/bash arch/arm64/
If you fix the issue, kindly add following tag where applicable
All warnings (new ones prefixed by >>):
>> arch/arm64/kvm/hyp/nvhe/stacktrace.c:46:6: warning: no previous prototype for function 'kvm_nvhe_prepare_backtrace' [-Wmissing-prototypes]
void kvm_nvhe_prepare_backtrace(unsigned long fp, unsigned long pc)
^ arch/arm64/kvm/hyp/nvhe/stacktrace.c:46:1: note: declare 'static' if the function is not intended to be used outside of this translation unit void kvm_nvhe_prepare_backtrace(unsigned long fp, unsigned long pc)
^
static
1 warning generated.
vim +/kvm_nvhe_prepare_backtrace +46 arch/arm64/kvm/hyp/nvhe/stacktrace.c
36
37 /*
38 * kvm_nvhe_prepare_backtrace - prepare to dump the nVHE backtrace
39 *
40 * @fp : frame pointer at which to start the unwinding.
41 * @pc : program counter at which to start the unwinding.
42 *
43 * Saves the information needed by the host to dump the nVHE hypervisor
44 * backtrace.
45 */
> 46 void kvm_nvhe_prepare_backtrace(unsigned long fp, unsigned long pc)Signature · 2 lines
0-DAY CI Kernel Test Service https://01.org/lkp
PATCH v6 10/17 KVM: arm64: Implement non-protected nVHE hyp stack unwinder
In reply to Kalesh SinghImplements the common framework necessary for unwind() to work for non-protected nVHE mode:
- on_accessible_stack()
- on_overflow_stack()
- unwind_next()
Non-protected nVHE unwind() is used to unwind and dump the hypervisor stacktrace by the host in EL1
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Use regular comments instead of doc comments, per Fuad
arch/arm64/include/asm/stacktrace/common.h | 2 + arch/arm64/include/asm/stacktrace/nvhe.h | 76 +++++++++++++++++++++- arch/arm64/kvm/arm.c | 2 +- 3 files changed, 77 insertions(+), 3 deletions(-)
arch/arm64/include/asm/stacktrace/common.h +2 -0
@@ -34,6 +34,7 @@ enum stack_type {STACK_TYPE_OVERFLOW ,STACK_TYPE_SDEI_NORMAL ,STACK_TYPE_SDEI_CRITICAL , +STACK_TYPE_HYP , __NR_STACK_TYPES };
@@ -186,6 +187,7 @@ static inline int unwind_next_common(struct unwind_state *state,* * TASK -> IRQ -> OVERFLOW -> SDEI_NORMAL * TASK -> SDEI_NORMAL -> SDEI_CRITICAL -> OVERFLOW +* HYP -> OVERFLOW * * ... but the nesting itself is strict. Once we transition from one * stack to another, it's never valid to unwind back to that first
arch/arm64/include/asm/stacktrace/nvhe.h +74 -2
99 lines not on this page · raw message
arch/arm64/kvm/arm.c +1 -1
@@ -49,7 +49,7 @@ DEFINE_STATIC_KEY_FALSE(kvm_protected_mode_initialized); DECLARE_KVM_HYP_PER_CPU(unsigned long, kvm_hyp_vector); -static DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page); +DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);unsigned long kvm_arm_hyp_percpu_base[NR_CPUS ]; DECLARE_KVM_NVHE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 11/17 KVM: arm64: Introduce hyp_dump_backtrace()
In reply to Kalesh SinghIn non-protected nVHE mode, unwinds and dumps the hypervisor backtrace from EL1. This is possible beacause the host can directly access the hypervisor stack pages in non-protected mode.
The nVHE backtrace is dumped on hyp_panic(), before panicking the host.
[ 101.498183] kvm [377]: nVHE call trace: [ 101.498363] kvm [377]: [<ffff8000090a6570>] __kvm_nvhe_hyp_panic+0xac/0xf8 [ 101.499045] kvm [377]: [<ffff8000090a65cc>] __kvm_nvhe_hyp_panic_bad_stack+0x10/0x10 [ 101.499498] kvm [377]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 . . . [ 101.524929] kvm [377]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 101.525062] kvm [377]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 101.525195] kvm [377]: [<ffff8000090a5de4>] __kvm_nvhe___kvm_vcpu_run+0x30/0x40c [ 101.525333] kvm [377]: [<ffff8000090a8b64>] __kvm_nvhe_handle___kvm_vcpu_run+0x30/0x48 [ 101.525468] kvm [377]: [<ffff8000090a88b8>] __kvm_nvhe_handle_trap+0xc4/0x128 [ 101.525602] kvm [377]: [<ffff8000090a7864>] __kvm_nvhe___host_exit+0x64/0x64 [ 101.525745] kvm [377]: ---[ end nVHE call trace ]---
---
Changes in v6:
- Fix some typos in commit text and comments, per Fuad
- Remove kvm_nvhe_print_backtrace_entry(), per Oliver
- To make nVHE call trace delimiters consistent between protected and non-protected mode, factor it out into helpers, per Oliver
- Change end delimiter to more match that of arm64 stacktrace
(---[ end nVHE call trace ]---), per Oliver
Changes in v5:
- Move code out from nvhe.h header to handle_exit.c, per Marc
- Fix stacktrace symbolization when CONFIG_RAMDOMIZE_BASE is enabled, per Fuad
- Use regular comments instead of doc comments, per Fuad
arch/arm64/include/asm/stacktrace/nvhe.h | 17 ++++++ arch/arm64/kvm/handle_exit.c | 69 ++++++++++++++++++++++++ 2 files changed, 86 insertions(+)
arch/arm64/include/asm/stacktrace/nvhe.h +17 -0
@@ -16,6 +16,23 @@#include <asm/stacktrace/common.h> +/* +* kvm_nvhe_unwind_init - Start an unwind from the given nVHE HYP fp and pc +* +* @state : unwind_state to initialize +* @fp : frame pointer at which to start the unwinding. +* @pc : program counter at which to start the unwinding. +*/ +static inline void kvm_nvhe_unwind_init (struct unwind_state *state, +unsigned long fp, +unsigned long pc) +{ +unwind_init_common (state,NULL ); + + state->fp = fp; + state->pc = pc; +} +static inline bool on_hyp_stack (unsigned long sp,unsigned long size,struct stack_info *info);
arch/arm64/kvm/handle_exit.c +69 -0
87 lines not on this page · raw message
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 12/17 KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig
In reply to Kalesh SinghThis can be used to disable stacktrace for the protected KVM nVHE hypervisor, in order to save on the associated memory usage.
This option is disabled by default, since protected KVM is not widely used on platforms other than Android currently.
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Make PROTECTED_NVHE_STACKTRACE depend on NVHE_EL2_DEBUG, per Marc
arch/arm64/kvm/Kconfig | 15 +++++++++++++++ 1 file changed, 15 insertions(+)
arch/arm64/kvm/Kconfig +15 -0
@@ -46,6 +46,21 @@ menuconfig KVM If unsure, say N. +config PROTECTED_NVHE_STACKTRACE + bool "Protected KVM hypervisor stacktraces" + depends on NVHE_EL2_DEBUG + default n + help + Say Y here to enable pKVM hypervisor stacktraces on hyp_panic() + + If you are not using protected nVHE (pKVM), say N. + + If using protected nVHE mode, but cannot afford the associated + memory cost (less than 0.75 page per CPU) of pKVM stacktraces, + say N. + + If unsure, say N. + config NVHE_EL2_DEBUG bool "Debug mode for non-VHE EL2 object" depends on KVM
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v6 12/17] KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig
Quoted 37 lines · Kalesh Singh
This can be used to disable stacktrace for the protected KVM nVHE hypervisor, in order to save on the associated memory usage.
This option is disabled by default, since protected KVM is not widely used on platforms other than Android currently.
Signed-off-by Kalesh Singh <[email protected]>Reviewed-by Fuad Tabba <[email protected]>Tested-by Fuad Tabba <[email protected]>---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tagsChanges in v5:
- MakePROTECTED_NVHE_STACKTRACEdepend onNVHE_EL2_DEBUG, per Marcarch/arm64/kvm/Kconfig | 15 +++++++++++++++ 1 file changed, 15 insertions(+)arch/arm64/kvm/Kconfig +8
-0@@ -46,6 +46,21 @@ menuconfig KVM If unsure, say N. +config PROTECTED_NVHE_STACKTRACE + bool "Protected KVM hypervisor stacktraces" + depends on NVHE_EL2_DEBUG + default n + help + Say Y here to enable pKVM hypervisor stacktraces on hyp_panic() + + If you are not using protected nVHE (pKVM), say N.
arch/arm64/kvm/Kconfig menuconfig KVM · line 66
Quoted 10 lines
arch/arm64/kvm/Kconfig +7
-0+ + If using protected nVHE mode, but cannot afford the associated + memory cost (less than 0.75 page per CPU) of pKVM stacktraces, + say N. + + If unsure, say N. + config NVHE_EL2_DEBUG bool "Debug mode for non-VHE EL2 object" depends on KVM
Having NVHE_EL2_DEBUG after PROTECTED_NVHE_STACKTRACE is pretty odd when you use 'make menuconfig', as the new option appears out of sequence.
I'll move it around, no need to respin for this.
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 12/17] KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig
Quoted 47 lines · Marc Zyngier
Quoted 44 lines · Kalesh Singh
This can be used to disable stacktrace for the protected KVM nVHE hypervisor, in order to save on the associated memory usage.
This option is disabled by default, since protected KVM is not widely used on platforms other than Android currently.
Signed-off-by Kalesh Singh <[email protected]>Reviewed-by Fuad Tabba <[email protected]>Tested-by Fuad Tabba <[email protected]>---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tagsChanges in v5:
- MakePROTECTED_NVHE_STACKTRACEdepend onNVHE_EL2_DEBUG, per Marcarch/arm64/kvm/Kconfig | 15 +++++++++++++++ 1 file changed, 15 insertions(+)arch/arm64/kvm/Kconfig +14
-0@@ -46,6 +46,21 @@ menuconfig KVM If unsure, say N. +config PROTECTED_NVHE_STACKTRACE + bool "Protected KVM hypervisor stacktraces" + depends on NVHE_EL2_DEBUG + default n + help + Say Y here to enable pKVM hypervisor stacktraces on hyp_panic() + + If you are not using protected nVHE (pKVM), say N. + + If using protected nVHE mode, but cannot afford the associated + memory cost (less than 0.75 page per CPU) of pKVM stacktraces, + say N. + + If unsure, say N.
Quoted 8 lines
arch/arm64/kvm/Kconfig menuconfig KVM · line 66
Quoted 4 lines
arch/arm64/kvm/Kconfig +1
-0+ config NVHE_EL2_DEBUG bool "Debug mode for non-VHE EL2 object" depends on KVMHaving
NVHE_EL2_DEBUGafterPROTECTED_NVHE_STACKTRACEis pretty odd when you use 'make menuconfig', as the new option appears out of sequence.
Ahh good catch. Should have moved it when the dependency got changed.
Quoted 3 lines
I'll move it around, no need to respin for this.
Thanks Marc.
Quoted 6 lines
Thanks,
M.--
Without deviation from the norm, progress is not possible.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 13/17 KVM: arm64: Allocate shared pKVM hyp stacktrace buffers
In reply to Kalesh SinghIn protected nVHE mode the host cannot directly access hypervisor memory, so we will dump the hypervisor stacktrace to a shared buffer with the host.
The minimum size for the buffer required, assuming the min frame size of [x29, x30] (2 * sizeof(long)), is half the combined size of the hypervisor and overflow stacks plus an additional entry to delimit the end of the stacktrace.
The stacktrace buffers are used later in the series to dump the nVHE hypervisor stacktrace when using protected-mode.
---
Changes in v6:
- Fix typo in commit text, per Fuad
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Fix typo in commit text, per Marc
arch/arm64/include/asm/memory.h | 8 ++++++++ arch/arm64/kvm/hyp/nvhe/stacktrace.c | 4 ++++ 2 files changed, 12 insertions(+)
arch/arm64/include/asm/memory.h +8 -0
@@ -113,6 +113,14 @@#define OVERFLOW_STACK_SIZE SZ_4K +/* +* With the minimum frame size of [x29, x30], exactly half the combined +* sizes of the hyp and overflow stacks is the maximum size needed to +* save the unwinded stacktrace; plus an additional entry to delimit the +* end. +*/ +#define NVHE_STACKTRACE_SIZE ((OVERFLOW_STACK_SIZE + PAGE_SIZE) / 2 + sizeof(long)) +/* * Alignment of kernel segments (e.g. .text, .data). *
arch/arm64/kvm/hyp/nvhe/stacktrace.c +4 -0
@@ -34,6 +34,10 @@ static void hyp_prepare_backtrace(unsigned long fp, unsigned long pc) stacktrace_info->pc = pc; } +#ifdef CONFIG_PROTECTED_NVHE_STACKTRACE +DEFINE_PER_CPU(unsigned long [NVHE_STACKTRACE_SIZE/sizeof(long)], pkvm_stacktrace); +#endif /* CONFIG_PROTECTED_NVHE_STACKTRACE */ +/* * kvm_nvhe_prepare_backtrace - prepare to dump the nVHE backtrace *
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 14/17 KVM: arm64: Stub implementation of pKVM HYP stack unwinder
In reply to Kalesh SinghAdd some stub implementations of protected nVHE stack unwinder, for building. These are implemented later in this series.
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
Changes in v5:
- Mark unwind_next() as inline, per Marc
arch/arm64/include/asm/stacktrace/nvhe.h | 35 ++++++++++++++++++++++-- 1 file changed, 33 insertions(+), 2 deletions(-)
arch/arm64/include/asm/stacktrace/nvhe.h +33 -2
@@ -9,6 +9,10 @@* to allocate shared buffers for the host to read the unwinded * stacktrace. * +* 2) pKVM (protected nVHE) mode - the host cannot directly access +* the HYP memory. The stack is unwinded in EL2 and dumped to a shared +* buffer where the host can read and print the stacktrace. +* * Copyright (C) 2022 Google LLC */ #ifndef __ASM_STACKTRACE_NVHE_H
@@ -49,7 +53,34 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,return false; } -#ifndef __KVM_NVHE_HYPERVISOR__ +#ifdef __KVM_NVHE_HYPERVISOR__ +/* +* Protected nVHE HYP stack unwinder +* +* In protected mode, the unwinding is done by the hypervisor in EL2. +*/ + +#ifdef CONFIG_PROTECTED_NVHE_STACKTRACE +static inline bool on_overflow_stack (unsigned long sp,unsigned long size, +struct stack_info *info) +{ +return false; +} + +static inline bool on_hyp_stack (unsigned long sp,unsigned long size, +struct stack_info *info) +{ +return false; +} + +static inline int notrace unwind_next(struct unwind_state *state) +{ +return 0 ; +} +NOKPROBE_SYMBOL (unwind_next); +#endif /* CONFIG_PROTECTED_NVHE_STACKTRACE */ + +#else /* !__KVM_NVHE_HYPERVISOR__ */ /* * Conventional (non-protected) nVHE HYP stack unwinder *
@@ -132,5 +163,5 @@ static inline int notrace unwind_next(struct unwind_state *state) }NOKPROBE_SYMBOL (unwind_next); -#endif /* !__KVM_NVHE_HYPERVISOR__ */ +#endif /* __KVM_NVHE_HYPERVISOR__ */ #endif /* __ASM_STACKTRACE_NVHE_H */
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 15/17 KVM: arm64: Save protected-nVHE (pKVM) hyp stacktrace
In reply to Kalesh SinghIn protected nVHE mode, the host cannot access private owned hypervisor memory. Also the hypervisor aims to remains simple to reduce the attack surface and does not provide any printk support.
For the above reasons, the approach taken to provide hypervisor stacktraces in protected mode is:
1) Unwind and save the hyp stack addresses in EL2 to a shared buffer with the host (done in this patch).
2) Delegate the dumping and symbolization of the addresses to the host in EL1 (later patch in the series).
On hyp_panic(), the hypervisor prepares the stacktrace before returning to the host.
---
Changes in v6:
- Simplify pkvm_save_backtrace_entry() using array semantics instead of the pointer arithmetic, per Oliver.
Changes in v5:
- Comment/clarify pkvm_save_backtrace_entry(), per Fuad
- kvm_nvhe_unwind_init(), doesn't need to be always inline, make it inline instead to avoid linking issues, per Marc
- Use regular comments instead of doc comments, per Fuad
arch/arm64/kvm/hyp/nvhe/stacktrace.c | 55 +++++++++++++++++++++++++++- 1 file changed, 54 insertions(+), 1 deletion(-)
arch/arm64/kvm/hyp/nvhe/stacktrace.c +54 -1
@@ -35,7 +35,60 @@ static void hyp_prepare_backtrace(unsigned long fp, unsigned long pc) }#ifdef CONFIG_PROTECTED_NVHE_STACKTRACE +#include <asm/stacktrace/nvhe.h> + DEFINE_PER_CPU(unsigned long [NVHE_STACKTRACE_SIZE/sizeof(long)], pkvm_stacktrace); + +/* +* pkvm_save_backtrace_entry - Saves a protected nVHE HYP stacktrace entry +* +* @arg : index of the entry in the stacktrace buffer +* @where : the program counter corresponding to the stack frame +* +* Save the return address of a stack frame to the shared stacktrace buffer. +* The host can access this shared buffer from EL1 to dump the backtrace. +*/ +static bool pkvm_save_backtrace_entry (void *arg,unsigned long where) +{ +unsigned long *stacktrace =this_cpu_ptr (pkvm_stacktrace); +int size =NVHE_STACKTRACE_SIZE /sizeof (long ); +int *idx = (int *)arg; + +/* +* Need 2 free slots: 1 for current entry and 1 for the +* delimiter. +*/ +if (*idx > size -2 ) +return false; + + stacktrace[*idx] = where; + stacktrace[++*idx] =0UL ; + +return true; +} + +/* +* pkvm_save_backtrace - Saves the protected nVHE HYP stacktrace +* +* @fp : frame pointer at which to start the unwinding. +* @pc : program counter at which to start the unwinding. +* +* Save the unwinded stack addresses to the shared stacktrace buffer. +* The host can access this shared buffer from EL1 to dump the backtrace. +*/ +static void pkvm_save_backtrace (unsigned long fp,unsigned long pc) +{ +struct unwind_state state; +int idx =0 ; + +kvm_nvhe_unwind_init (&state, fp, pc); + +unwind (&state, pkvm_save_backtrace_entry, &idx); +} +#else /* !CONFIG_PROTECTED_NVHE_STACKTRACE */ +static void pkvm_save_backtrace (unsigned long fp,unsigned long pc) +{ +}#endif /* CONFIG_PROTECTED_NVHE_STACKTRACE */ /*
@@ -50,7 +103,7 @@ DEFINE_PER_CPU(unsigned long [NVHE_STACKTRACE_SIZE/sizeof(long)], pkvm_stacktracvoid kvm_nvhe_prepare_backtrace (unsigned long fp,unsigned long pc) {if (is_protected_kvm_enabled ()) -return ; +pkvm_save_backtrace (fp, pc);else hyp_prepare_backtrace (fp, pc); }
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 16/17 KVM: arm64: Implement protected nVHE hyp stack unwinder
In reply to Kalesh SinghImplements the common framework necessary for unwind() to work in the protected nVHE context:
- on_accessible_stack()
- on_overflow_stack()
- unwind_next()
Protected nVHE unwind() is used to unwind and save the hyp stack addresses to the shared stacktrace buffer. The host reads the entries in this buffer, symbolizes and dumps the stacktrace (later patch in the series).
---
Changes in v6:
- Add Fuad’s Reviewed-by and Tested-by tags
arch/arm64/include/asm/stacktrace/nvhe.h | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-)
arch/arm64/include/asm/stacktrace/nvhe.h +12 -3
@@ -64,18 +64,27 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,static inline bool on_overflow_stack (unsigned long sp,unsigned long size,struct stack_info *info) { -return false; +unsigned long low = (unsigned long )this_cpu_ptr (overflow_stack); +unsigned long high = low +OVERFLOW_STACK_SIZE ; + +return on_stack (sp, size, low, high,STACK_TYPE_OVERFLOW , info); }static inline bool on_hyp_stack (unsigned long sp,unsigned long size,struct stack_info *info) { -return false; +struct kvm_nvhe_init_params *params =this_cpu_ptr (&kvm_init_params); +unsigned long high = params->stack_hyp_va; +unsigned long low = high -PAGE_SIZE ; + +return on_stack (sp, size, low, high,STACK_TYPE_HYP , info); } static inline int notrace unwind_next(struct unwind_state *state) { -return 0 ; +struct stack_info info; + +return unwind_next_common (state, &info,NULL ); }NOKPROBE_SYMBOL (unwind_next);#endif /* CONFIG_PROTECTED_NVHE_STACKTRACE */
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 17/17 KVM: arm64: Introduce pkvm_dump_backtrace()
In reply to Kalesh SinghDumps the pKVM hypervisor backtrace from EL1 by reading the unwinded addresses from the shared stacktrace buffer.
The nVHE hyp backtrace is dumped on hyp_panic(), before panicking the host.
[ 111.623091] kvm [367]: nVHE call trace: [ 111.623215] kvm [367]: [<ffff8000090a6570>] __kvm_nvhe_hyp_panic+0xac/0xf8 [ 111.623448] kvm [367]: [<ffff8000090a65cc>] __kvm_nvhe_hyp_panic_bad_stack+0x10/0x10 [ 111.623642] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 . . . [ 111.640366] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 111.640467] kvm [367]: [<ffff8000090a61e4>] __kvm_nvhe_recursive_death+0x24/0x34 [ 111.640574] kvm [367]: [<ffff8000090a5de4>] __kvm_nvhe___kvm_vcpu_run+0x30/0x40c [ 111.640676] kvm [367]: [<ffff8000090a8b64>] __kvm_nvhe_handle___kvm_vcpu_run+0x30/0x48 [ 111.640778] kvm [367]: [<ffff8000090a88b8>] __kvm_nvhe_handle_trap+0xc4/0x128 [ 111.640880] kvm [367]: [<ffff8000090a7864>] __kvm_nvhe___host_exit+0x64/0x64 [ 111.640996] kvm [367]: ---[ end nVHE call trace ]---
---
Changes in v6:
- And range check when dumping pkvm stacktrace, per Oliver
- Use consistent nVHE call trace delimiters between protected and non-protected mode, per Oliver
- Fix typo in comment, per Fuad
Changes in v5:
- Move code out from nvhe.h header to handle_exit.c, per Marc
- Fix stacktrace symbolization when CONFIG_RAMDOMIZE_BASE is enabled, per Fuad
- Use regular comments instead of doc comments, per Fuad
arch/arm64/kvm/handle_exit.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)
arch/arm64/kvm/handle_exit.c +34 -1
@@ -371,6 +371,39 @@ static void hyp_dump_backtrace(unsigned long hyp_offset)kvm_nvhe_dump_backtrace_end (); } +#ifdef CONFIG_PROTECTED_NVHE_STACKTRACE +DECLARE_KVM_NVHE_PER_CPU(unsigned long [NVHE_STACKTRACE_SIZE/sizeof(long)], + pkvm_stacktrace); + +/* +* pkvm_dump_backtrace - Dump the protected nVHE HYP backtrace. +* +* @hyp_offset: hypervisor offset, used for address translation. +* +* Dumping of the pKVM HYP backtrace is done by reading the +* stack addresses from the shared stacktrace buffer, since the +* host cannot directly access hypervisor memory in protected +* mode. +*/ +static void pkvm_dump_backtrace (unsigned long hyp_offset) +{ +unsigned long *stacktrace + = (unsigned long *)this_cpu_ptr_nvhe_sym (pkvm_stacktrace); +int i, size =NVHE_STACKTRACE_SIZE /sizeof (long ); + +kvm_nvhe_dump_backtrace_start (); +/* The saved stacktrace is terminated by a null entry */ +for (i =0 ; i < size && stacktrace[i]; i++) +kvm_nvhe_dump_backtrace_entry ((void *)hyp_offset, stacktrace[i]); +kvm_nvhe_dump_backtrace_end (); +} +#else /* !CONFIG_PROTECTED_NVHE_STACKTRACE */ +static void pkvm_dump_backtrace (unsigned long hyp_offset) +{ +kvm_err ("Cannot dump pKVM nVHE stacktrace: !CONFIG_PROTECTED_NVHE_STACKTRACE\n" ); +} +#endif /* CONFIG_PROTECTED_NVHE_STACKTRACE */ +/* * kvm_nvhe_dump_backtrace - Dump KVM nVHE hypervisor backtrace. *
@@ -379,7 +412,7 @@ static void hyp_dump_backtrace(unsigned long hyp_offset)static void kvm_nvhe_dump_backtrace (unsigned long hyp_offset) {if (is_protected_kvm_enabled ()) -return ; +pkvm_dump_backtrace (hyp_offset);else hyp_dump_backtrace (hyp_offset); }
Signature · 6 lines
2.37.1.359.gd136c6c3e2-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainercover KVM: arm64: nVHE stack unwinder rework
Hi all,
As Kalesh's series[1] already went through quite a few rounds and that it has proved to be an extremely useful debugging help, I'd like to queue it for 5.20.
However, there is a couple of nits that I'd like to address:
- the code is extremely hard to follow, due to the include maze and the various levels of inline functions that have forward declarations...
- there is a subtle bug in the way the kernel on_accessible_stack() helper has been rewritten
- the config symbol for the protected unwinder is oddly placed
Instead of going for another round and missing the merge window, I propose to stash the following patches on top, which IMHO result in something much more readable.
This series directly applies on top of Kalesh's.
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (5):
KVM: arm64: Move PROTECTED_NVHE_STACKTRACE around
KVM: arm64: Move nVHE stacktrace unwinding into its own compilation
unit
KVM: arm64: Make unwind()/on_accessible_stack() per-unwinder functions
KVM: arm64: Move nVHE-only helpers into kvm/stacktrace.c
arm64: Update 'unwinder howto'
Oliver Upton (1):
KVM: arm64: Don't open code ARRAY_SIZE()
arch/arm64/include/asm/stacktrace.h | 74 ------- arch/arm64/include/asm/stacktrace/common.h | 69 ++----- arch/arm64/include/asm/stacktrace/nvhe.h | 125 +----------- arch/arm64/kernel/stacktrace.c | 90 +++++++++ arch/arm64/kvm/Kconfig | 24 ++- arch/arm64/kvm/Makefile | 2 +- arch/arm64/kvm/handle_exit.c | 98 --------- arch/arm64/kvm/hyp/nvhe/stacktrace.c | 55 +++++- arch/arm64/kvm/stacktrace.c | 218 +++++++++++++++++++++ 9 files changed, 394 insertions(+), 361 deletions(-) create mode 100644 arch/arm64/kvm/stacktrace.c
Signature · 1 line
2.34.1
Marc Zyngier
maintainerPATCH 1/6 KVM: arm64: Move PROTECTED_NVHE_STACKTRACE around
Make the dependency with EL2_DEBUG more obvious by moving the stacktrace configurtion *after* it.
--- arch/arm64/kvm/Kconfig | 24 +++++++++++------------- 1 file changed, 11 insertions(+), 13 deletions(-)
arch/arm64/kvm/Kconfig +11 -13
@@ -46,6 +46,16 @@ menuconfig KVM If unsure, say N. +config NVHE_EL2_DEBUG + bool "Debug mode for non-VHE EL2 object" + depends on KVM + help + Say Y here to enable the debug mode for the non-VHE KVM EL2 object. + Failure reports will BUG() in the hypervisor. This is intended for + local EL2 hypervisor development. + + If unsure, say N. + config PROTECTED_NVHE_STACKTRACE bool "Protected KVM hypervisor stacktraces" depends on NVHE_EL2_DEBUG
@@ -53,22 +63,10 @@ config PROTECTED_NVHE_STACKTRACE help Say Y here to enable pKVM hypervisor stacktraces on hyp_panic() - If you are not using protected nVHE (pKVM), say N. - If using protected nVHE mode, but cannot afford the associated memory cost (less than 0.75 page per CPU) of pKVM stacktraces, say N. - If unsure, say N. - -config NVHE_EL2_DEBUG - bool "Debug mode for non-VHE EL2 object" - depends on KVM - help - Say Y here to enable the debug mode for the non-VHE KVM EL2 object. - Failure reports will BUG() in the hypervisor. This is intended for - local EL2 hypervisor development. - - If unsure, say N. + If unsure, or not using protected nVHE (pKVM), say N. endif # VIRTUALIZATION
Signature · 1 line
2.34.1
Marc Zyngier
maintainerPATCH 2/6 KVM: arm64: Move nVHE stacktrace unwinding into its own compilation unit
In reply to Marc ZyngierThe unwinding code doesn't really belong to the exit handling code. Instead, move it to a file (conveniently named stacktrace.c to confuse the reviewer), and move all the stacktrace-related stuff there.
It will be joined by more code very soon.
--- arch/arm64/include/asm/stacktrace/nvhe.h | 2 + arch/arm64/kvm/Makefile | 2 +- arch/arm64/kvm/handle_exit.c | 98 ------------------ arch/arm64/kvm/stacktrace.c | 120 +++++++++++++++++++++++ 4 files changed, 123 insertions(+), 99 deletions(-) create mode 100644 arch/arm64/kvm/stacktrace.c
arch/arm64/include/asm/stacktrace/nvhe.h +2 -0
@@ -172,5 +172,7 @@ static inline int notrace unwind_next(struct unwind_state *state) }NOKPROBE_SYMBOL (unwind_next); +void kvm_nvhe_dump_backtrace (unsigned long hyp_offset); +#endif /* __KVM_NVHE_HYPERVISOR__ */ #endif /* __ASM_STACKTRACE_NVHE_H */
arch/arm64/kvm/Makefile +1 -1
@@ -12,7 +12,7 @@ obj-$(CONFIG_KVM) += hyp/ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o pvtime.o \ inject_fault.o va_layout.o handle_exit.o \ - guest.o debug.o reset.o sys_regs.o \ + guest.o debug.o reset.o sys_regs.o stacktrace.o \ vgic-sys-reg-v3.o fpsimd.o pkvm.o \ arch_timer.o trng.o vmid.o \ vgic/vgic.o vgic/vgic-init.o \
arch/arm64/kvm/handle_exit.c +0 -98
104 lines not on this page · raw message
arch/arm64/kvm/stacktrace.c new file +120 -0
120 lines not on this page · raw message
Signature · 1 line
2.34.1
Marc Zyngier
maintainerPATCH 3/6 KVM: arm64: Make unwind()/on_accessible_stack() per-unwinder functions
In reply to Marc ZyngierHaving multiple versions of on_accessible_stack() (one per unwinder) makes it very hard to reason about what is used where due to the complexity of the various includes, the forward declarations, and the reliance on everything being 'inline'.
Instead, move the code back where it should be. Each unwinder implements:
- on_accessible_stack() as well as the helpers it depends on,
- unwind()/unwind_next(), as they pass on_accessible_stack as a parameter to unwind_next_common() (which is the only common code here)
This hardly results in any duplication, and makes it much easier to reason about the code.
--- arch/arm64/include/asm/stacktrace.h | 74 ------------------ arch/arm64/include/asm/stacktrace/common.h | 55 ++++--------- arch/arm64/include/asm/stacktrace/nvhe.h | 84 +------------------- arch/arm64/kernel/stacktrace.c | 90 ++++++++++++++++++++++ arch/arm64/kvm/hyp/nvhe/stacktrace.c | 52 +++++++++++++ arch/arm64/kvm/stacktrace.c | 55 +++++++++++++ 6 files changed, 213 insertions(+), 197 deletions(-)
arch/arm64/include/asm/stacktrace.h +0 -74
@@ -57,78 +57,4 @@ static inline bool on_overflow_stack(unsigned long sp, unsigned long size, struct stack_info *info) { return false; }#endif - -/* -* We can only safely access per-cpu stacks from current in a non-preemptible -* context. -*/ -static inline bool on_accessible_stack (const struct task_struct *tsk, -unsigned long sp,unsigned long size, -struct stack_info *info) -{ -if (on_accessible_stack_common (tsk, sp, size, info)) -return true; - -if (on_task_stack (tsk, sp, size, info)) -return true; -if (tsk != current ||preemptible ()) -return false; -if (on_irq_stack (sp, size, info)) -return true; -if (on_sdei_stack (sp, size, info)) -return true; - -return false; -} - -/* -* Unwind from one frame record (A) to the next frame record (B). -* -* We terminate early if the location of B indicates a malformed chain of frame -* records (e.g. a cycle), determined based on the location and fp value of A -* and the location (but not the fp value) of B. -*/ -static inline int notrace unwind_next(struct unwind_state *state) -{ -struct task_struct *tsk = state->task; -unsigned long fp = state->fp; -struct stack_info info; -int err; - -/* Final frame; nothing to unwind */ -if (fp == (unsigned long )task_pt_regs (tsk)->stackframe) -return -ENOENT ; - - err =unwind_next_common (state, &info,NULL ); -if (err) -return err; - - state->pc =ptrauth_strip_insn_pac (state->pc); - -#ifdef CONFIG_FUNCTION_GRAPH_TRACER -if (tsk->ret_stack && - (state->pc == (unsigned long )return_to_handler)) { -unsigned long orig_pc; -/* -* This is a case where function graph tracer has -* modified a return address (LR) in a stack frame -* to hook a function return. -* So replace it to an original value. -*/ - orig_pc =ftrace_graph_ret_addr (tsk,NULL , state->pc, - (void *)state->fp); -if (WARN_ON_ONCE (state->pc == orig_pc)) -return -EINVAL ; - state->pc = orig_pc; - } -#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ -#ifdef CONFIG_KRETPROBES -if (is_kretprobe_trampoline (state->pc)) - state->pc =kretprobe_find_ret_addr (tsk, (void *)state->fp, &state->kr_cur); -#endif - -return 0 ; -} -NOKPROBE_SYMBOL (unwind_next); -#endif /* __ASM_STACKTRACE_H */
arch/arm64/include/asm/stacktrace/common.h +15 -40
85 lines not on this page · raw message
arch/arm64/include/asm/stacktrace/nvhe.h +1 -83
96 lines not on this page · raw message
arch/arm64/kernel/stacktrace.c +90 -0
96 lines not on this page · raw message
arch/arm64/kvm/hyp/nvhe/stacktrace.c +52 -0
58 lines not on this page · raw message
arch/arm64/kvm/stacktrace.c +55 -0
61 lines not on this page · raw message
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 3/6] KVM: arm64: Make unwind()/on_accessible_stack() per-unwinder functions
Quoted 5 lines · Marc Zyngier
Having multiple versions of
on_accessible_stack()(one per unwinder) makes it very hard to reason about what is used where due to the complexity of the various includes, the forward declarations, and the reliance on everything being 'inline'.
Quoted 8 lines
Instead, move the code back where it should be. Each unwinder implements:
-
on_accessible_stack()as well as the helpers it depends on,- unwind()/unwind_next(), as they pass
on_accessible_stackas a parameter tounwind_next_common()(which is the only common code here)
It feels like more of the accessibility stuff *should* be sharable, but yeah.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 4/6 KVM: arm64: Move nVHE-only helpers into kvm/stacktrace.c
In reply to Marc Zyngierkvm_nvhe_stack_kern_va() only makes sense as part of the nVHE unwinder, so simply move it there.
--- arch/arm64/include/asm/stacktrace/nvhe.h | 41 ------------------------ arch/arm64/kvm/stacktrace.c | 41 ++++++++++++++++++++++++ 2 files changed, 41 insertions(+), 41 deletions(-)
arch/arm64/include/asm/stacktrace/nvhe.h +0 -41
@@ -49,47 +49,6 @@ DECLARE_KVM_NVHE_PER_CPU(unsigned long [OVERFLOW_STACK_SIZE/sizeof(long)], overf DECLARE_KVM_NVHE_PER_CPU(struct kvm_nvhe_stacktrace_info, kvm_stacktrace_info); DECLARE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page); -/* -* kvm_nvhe_stack_kern_va - Convert KVM nVHE HYP stack addresses to a kernel VAs -* -* The nVHE hypervisor stack is mapped in the flexible 'private' VA range, to -* allow for guard pages below the stack. Consequently, the fixed offset address -* translation macros won't work here. -* -* The kernel VA is calculated as an offset from the kernel VA of the hypervisor -* stack base. -* -* Returns true on success and updates @addr to its corresponding kernel VA; -* otherwise returns false. -*/ -static inline bool kvm_nvhe_stack_kern_va (unsigned long *addr, -enum stack_type type) -{ -struct kvm_nvhe_stacktrace_info *stacktrace_info; -unsigned long hyp_base, kern_base, hyp_offset; - - stacktrace_info =this_cpu_ptr_nvhe_sym (kvm_stacktrace_info); - -switch (type) { -case STACK_TYPE_HYP : - kern_base = (unsigned long )*this_cpu_ptr (&kvm_arm_hyp_stack_page); - hyp_base = (unsigned long )stacktrace_info->stack_base; -break ; -case STACK_TYPE_OVERFLOW : - kern_base = (unsigned long )this_cpu_ptr_nvhe_sym (overflow_stack); - hyp_base = (unsigned long )stacktrace_info->overflow_stack_base; -break ; -default : -return false; - } - - hyp_offset = *addr - hyp_base; - - *addr = kern_base + hyp_offset; - -return true; -} -void kvm_nvhe_dump_backtrace (unsigned long hyp_offset);#endif /* __KVM_NVHE_HYPERVISOR__ */
arch/arm64/kvm/stacktrace.c +41 -0
47 lines not on this page · raw message
Signature · 1 line
2.34.1
Marc Zyngier
maintainerPATCH 5/6 KVM: arm64: Don't open code ARRAY_SIZE()
In reply to Marc ZyngierFrom: Oliver Upton <[email protected]>
Use ARRAY_SIZE() instead of an open-coded version.
--- arch/arm64/kvm/hyp/nvhe/stacktrace.c | 3 +-- arch/arm64/kvm/stacktrace.c | 6 ++++-- 2 files changed, 5 insertions(+), 4 deletions(-)
arch/arm64/kvm/hyp/nvhe/stacktrace.c +1 -2
@@ -103,14 +103,13 @@ static void notrace unwind(struct unwind_state *state,static bool pkvm_save_backtrace_entry (void *arg,unsigned long where) {unsigned long *stacktrace =this_cpu_ptr (pkvm_stacktrace); -int size =NVHE_STACKTRACE_SIZE /sizeof (long );int *idx = (int *)arg;/* * Need 2 free slots: 1 for current entry and 1 for the * delimiter. */ -if (*idx > size -2 ) +if (*idx >ARRAY_SIZE (pkvm_stacktrace) -2 )return false; stacktrace[*idx] = where;
arch/arm64/kvm/stacktrace.c +4 -2
@@ -187,11 +187,13 @@ static void pkvm_dump_backtrace(unsigned long hyp_offset) {unsigned long *stacktrace = (unsigned long *)this_cpu_ptr_nvhe_sym (pkvm_stacktrace); -int i, size =NVHE_STACKTRACE_SIZE /sizeof (long ); +int i;kvm_nvhe_dump_backtrace_start ();/* The saved stacktrace is terminated by a null entry */ -for (i =0 ; i < size && stacktrace[i]; i++) +for (i =0 ; + i <ARRAY_SIZE (kvm_nvhe_sym (pkvm_stacktrace)) && stacktrace[i]; + i++)kvm_nvhe_dump_backtrace_entry ((void *)hyp_offset, stacktrace[i]);kvm_nvhe_dump_backtrace_end (); }
Signature · 1 line
2.34.1
Marc Zyngier
maintainerPATCH 6/6 arm64: Update 'unwinder howto'
In reply to Marc ZyngierImplementing a new unwinder is a bit more involved than writing a couple of helpers, so let's not lure the reader into a false sense of comfort. Instead, let's point out what they should call into, and what sort of parameter they need to provide.
--- arch/arm64/include/asm/stacktrace/common.h | 14 ++++---------- 1 file changed, 4 insertions(+), 10 deletions(-)
arch/arm64/include/asm/stacktrace/common.h +4 -10
@@ -5,17 +5,11 @@* To implement a new arm64 stack unwinder: * 1) Include this header * -* 2) Provide implementations for the following functions: -* on_overflow_stack(): Returns true if SP is on the overflow -* stack. -* on_accessible_stack(): Returns true is SP is on any accessible -* stack. -* unwind_next(): Performs validation checks on the frame -* pointer, and transitions unwind_state -* to the next frame. +* 2) Call into unwind_next_common() from your top level unwind +* function, passing it the validation and translation callbacks +* (though the later can be NULL if no translation is required). * -* See: arch/arm64/include/asm/stacktrace.h for reference -* implementations. +* See: arch/arm64/kernel/stacktrace.c for the reference implementation. * * Copyright (C) 2012 ARM Ltd. */
Signature · 1 line
2.34.1
Re: [PATCH 0/6] KVM: arm64: nVHE stack unwinder rework
In reply to Marc ZyngierQuoted 9 lines · Marc Zyngier
Hi all,
As Kalesh's series[1] already went through quite a few rounds and that it has proved to be an extremely useful debugging help, I'd like to queue it for 5.20.
However, there is a couple of nits that I'd like to address:
Quoted 8 lines
- the code is extremely hard to follow, due to the include maze and the various levels of inline functions that have forward declarations...
- there is a subtle bug in the way the kernel
on_accessible_stack()helper has been rewritten- the config symbol for the protected unwinder is oddly placed
Hi Marc,
Thanks for doing this rework.
For the series:
Thanks,
Kalesh
Quoted 35 lines
Instead of going for another round and missing the merge window, I propose to stash the following patches on top, which IMHO result in something much more readable.
This series directly applies on top of Kalesh's.
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (5):
KVM: arm64: MovePROTECTED_NVHE_STACKTRACEaround
KVM: arm64: Move nVHE stacktrace unwinding into its own compilation
unit
KVM: arm64: Make unwind()/on_accessible_stack() per-unwinder functions
KVM: arm64: Move nVHE-only helpers intokvm/stacktrace.c
arm64: Update 'unwinder howto'Oliver Upton (1):
KVM: arm64: Don't open codeARRAY_SIZE()arch/arm64/include/asm/stacktrace.h | 74 ------- arch/arm64/include/asm/stacktrace/common.h | 69 ++----- arch/arm64/include/asm/stacktrace/nvhe.h | 125 +----------- arch/arm64/kernel/stacktrace.c | 90 +++++++++ arch/arm64/kvm/Kconfig | 24 ++- arch/arm64/kvm/Makefile | 2 +- arch/arm64/kvm/handle_exit.c | 98 --------- arch/arm64/kvm/hyp/nvhe/stacktrace.c | 55 +++++- arch/arm64/kvm/stacktrace.c | 218 +++++++++++++++++++++ 9 files changed, 394 insertions(+), 361 deletions(-) create mode 100644 arch/arm64/kvm/stacktrace.c--
2.34.1
Re: [PATCH 0/6] KVM: arm64: nVHE stack unwinder rework
In reply to Marc ZyngierQuoted 17 lines · Marc Zyngier
Hi all,
As Kalesh's series[1] already went through quite a few rounds and that it has proved to be an extremely useful debugging help, I'd like to queue it for 5.20.
However, there is a couple of nits that I'd like to address:
- the code is extremely hard to follow, due to the include maze and the various levels of inline functions that have forward declarations...
- there is a subtle bug in the way the kernel
on_accessible_stack()helper has been rewritten- the config symbol for the protected unwinder is oddly placed
Quoted 7 lines
Instead of going for another round and missing the merge window, I propose to stash the following patches on top, which IMHO result in something much more readable.
This series directly applies on top of Kalesh's.
For the series (besides my own patch of course):
--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 0/6] KVM: arm64: nVHE stack unwinder rework
In reply to Marc ZyngierQuoted 11 lines · Marc Zyngier
As Kalesh's series[1] already went through quite a few rounds and that it has proved to be an extremely useful debugging help, I'd like to queue it for 5.20.
However, there is a couple of nits that I'd like to address:
- the code is extremely hard to follow, due to the include maze and the various levels of inline functions that have forward declarations...
[...]
Applied to next, thanks!
[1/6] KVM: arm64: Move PROTECTED_NVHE_STACKTRACE around
commit: 03fe9cd05b9f38353208c23bd791dac47c912054
[2/6] KVM: arm64: Move nVHE stacktrace unwinding into its own compilation unit
commit: 9f5fee05f6897d0fe0e3a44ade71bb85cd97b2ef
[3/6] KVM: arm64: Make unwind()/on_accessible_stack() per-unwinder functions
commit: 4e00532f37365967e9896966b1fe61888e659259
[4/6] KVM: arm64: Move nVHE-only helpers into kvm/stacktrace.c
commit: 0e773da1e688a1425ef7deae58fa11c5c7e09533
[5/6] KVM: arm64: Don't open code ARRAY_SIZE()
commit: 62ae21627aa96f6ef361981dd181c74dc7aa314c
[6/6] arm64: Update 'unwinder howto'
commit: a4c750e2328a117dc9b19a2a61db0d4347902029Cheers,
M.
Signature · 6 lines
Marc Zyngier <[email protected]> _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 11 lines · Kalesh Singh
This is v6 of nVHE stacktrace support. The series is based on arm64 for-next/stacktrace.
The previous versions were posted at:
v5: https://lore.kernel.org/r/[email protected]/
v4: https://lore.kernel.org/r/[email protected]/
v3: https://lore.kernel.org/r/[email protected]/
v2: https://lore.kernel.org/r/[email protected]/
v1: https://lore.kernel.org/r/[email protected]/[...]
Applied to next, thanks!
[01/17] arm64: stacktrace: Add shared header for common stack unwinding code
commit: 6bf212c89c48458d8deef1c973678c62528dab04
[02/17] arm64: stacktrace: Factor out on_accessible_stack_common()
commit: 15a59f19a015185bff90a68f601caec151dea4b4
[03/17] arm64: stacktrace: Factor out unwind_next_common()
commit: be63c647fd28d25484257f5f36a008db7d99991d
[04/17] arm64: stacktrace: Handle frame pointer from different address spaces
commit: 5b1b08619f50422c3e43d1fd7af257595a9e4a67
[05/17] arm64: stacktrace: Factor out common unwind()
commit: f51e7146740514347d6c5526a2c393e224a19c0d
[06/17] arm64: stacktrace: Add description of stacktrace/common.h
commit: 051ece6758cc10c2a6f1700ffe86d23fbb0b2553
[07/17] KVM: arm64: On stack overflow switch to hyp overflow_stack
commit: 548ec3336f323db56260b312c232ab37285f0284
[08/17] KVM: arm64: Stub implementation of non-protected nVHE HYP stack unwinder
commit: 573e1e8275f7167ddd533c6e4e0f500f8be4d974
[09/17] KVM: arm64: Prepare non-protected nVHE hypervisor stacktrace
commit: 879e5ac7b2e4db05799a905b5a07fc9e5dedf651
[10/17] KVM: arm64: Implement non-protected nVHE hyp stack unwinder
commit: db129d486ebdf4e3168282236f9d9008b42cac7e
[11/17] KVM: arm64: Introduce hyp_dump_backtrace()
commit: 314a61dc31845c233e47c53db3fe6f34284034f4
[12/17] KVM: arm64: Add PROTECTED_NVHE_STACKTRACE Kconfig
commit: 72adac1bd234002a65cef738e0eebfd6c2ce2e30
[13/17] KVM: arm64: Allocate shared pKVM hyp stacktrace buffers
commit: 6928bcc84bc4bd9a24a1cb1986418c3de76e1d99
[14/17] KVM: arm64: Stub implementation of pKVM HYP stack unwinder
commit: 25aa73b6db1831527cd4f14bf0ddf8dceadec802
[15/17] KVM: arm64: Save protected-nVHE (pKVM) hyp stacktrace
commit: 871c5d931417d3c0e1aa32c9e04da1dc74703843
[16/17] KVM: arm64: Implement protected nVHE hyp stack unwinder
commit: 75e9459e48d4867caf549e388bd4faabe1dbcbd3
[17/17] KVM: arm64: Introduce pkvm_dump_backtrace()
commit: 3a7e1b55aad45c0cf86bd4e2f212bb9a61905142Cheers,
M.
Signature · 6 lines
Marc Zyngier <[email protected]> _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm